sPDGFRβ and neuroinflammation are associated with AD biomarkers and differ by race: The ASCEND Study

Abstract INTRODUCTION There remains an urgent need to identify preclinical pathophysiological mechanisms of Alzheimer's disease (AD) development in high‐risk, racially diverse populations. We explored the relationship between cerebrospinal fluid (CSF) markers of vascular injury and neuroinflammation with AD biomarkers in middle‐aged Black/African American (B/AA) and non‐Hispanic White (NHW) participants. METHODS Adults (45–65 years) with a parental history of AD were enrolled (n = 82). CSF and blood biomarkers were collected at baseline and year 2. RESULTS CSF total tau (t‐tau), phosphorylated tau (p‐tau), and amyloid beta (Aβ)40 were elevated at year 2 compared to baseline. CSF soluble platelet‐derived growth factor receptor β (sPDGFRβ) levels, a marker of pericyte injury, correlated positively with t‐tau, p‐tau, Aβ40 markers of vascular injury, and cytokines at baseline and year 2. CSF sPDGFRβ and tau were significantly lower in B/AA than NHW. DISCUSSION Vascular dysfunction and neuroinflammation may precede cognitive decline and disease pathology in the very early preclinical stages of AD, and there are race‐related differences in these relationships. Highlights Cerebrospinal fluid (CSF) Alzheimer's disease (AD) biomarkers changed over 2 years in high‐risk middle‐aged adults. Markers of vascular dysfunction were associated with the CSF biomarkers amyloid beta and tau. AD biomarkers were lower in Black compared to non‐Hispanic White individuals. Markers of vascular dysfunction were lower among Black individuals.


INTRODUCTION
The increasing prevalence of Alzheimer's disease (AD) remains unaddressed due to the absence of effective disease-modifying therapies, and there remains a significant and urgent need to identify preclinical pathophysiological mechanisms responsible for AD development in high-risk, racially diverse populations.As pathophysiological mechanisms leading to the development of AD begin many years before clinical manifestations, 1 there is a need to study biomarkers that reflect underlying pathological processes that precede the emergence of clinical symptoms in AD. 1,2 Accumulation of these biomarkers, defined by amyloid beta (Aβ) deposition, formation of fibrillar tau, and neurodegeneration (AT[N]), is a progressive and continuous process that begins years before the onset of clinical symptoms, often during midlife. 2 Multifactorial putative mechanisms driving the neuropathology of AD, originally described by the amyloid cascade 3 and cholinergic 4 hypotheses, have led to the development of emerging complementary theories, including those that implicate vascular dysfunction, 5 angiotensin II, 6 and inflammation. 7Neurovascular dysfunction, including reduced cerebral blood flow and blood-brain barrier (BBB) breakdown, occur in early stages of AD. 8,9 Markers of endothelial injury are elevated in cerebrospinal fluid (CSF), serum, and pathological studies in AD. [10][11][12] Renin-angiotensin system (RAS) dysregulation is implicated in AD pathology. 6,13Increased angiotensin-converting enzyme (ACE1) activity and decreased ACE2 activity have been reported in CSF and post mortem brains in AD and are associated with AD pathology. 11,14Increased ACE1, proposed to play a role in Aβ degradation, is associated with higher levels of angiotensin II, which has multifactorial deleterious effects on several pathways associated with AD, including hypertension, vascular dysfunction, alterations in BBB function, and inflammation. 14,15Angiotensin II-induced cytokine release is a likely contributor to AD pathology through amplification of neuroinflammation and activation of microglia. 7,16,17ricytes appear to be crucial in maintaining vascular integrity and regulating cerebral blood flow and BBB permeability; 18,19 endothelial cell injury and damage to pericytes play a key role in AD pathology. 20,21Pericytes and endothelial-expressed matrix metalloproteinases (MMPs) and tissue inhibitor of metalloproteinases (TIMPs)   are involved in vascular homeostasis and dysregulated in AD. 22,23 In AD post mortem brain tissue, there is significant pericyte loss and a reduction in platelet-derived growth factor receptor β (PDGFRβ) levels. 24,25Pericytes in culture shed soluble PDGFRβ (sPDGFRβ) under hypoxic conditions or exposure to Aβ peptides. 26Elevated CSF levels of sPDGFRβ have been associated with BBB breakdown within the hippocampus in mildly cognitively impaired participants 15 and were reported to predict cognitive decline in early stages of AD independently of changes in Aβ and tau. 27We have previously shown that CSF sPDGFRβ levels were elevated from clinical AD cases and correlated with CSF albumin, a marker of BBB breakdown, and levels of CSF total tau (t-tau) and phosphorylated tau (p-tau). 28ack/African American adults (B/AAs) are 64% more likely to develop AD than non-Hispanic White adults (NHWs). 29The higher incidence and prevalence of AD among B/AAs are often attributed to sociocultural issues and biological and genetic factors related to higher cardiovascular risk, such as diabetes and hypertension.
More work is needed to examine race-associated differences in AD biomarkers as these differences may further contribute to health disparities in AD diagnosis and potential success of preventive interventions.
Recent studies in B/AA and NHW adults found that CSF levels of ttau and p-tau are lower in cognitively impaired B/AA older adults 30,31 and cognitively unimpaired B/AA middle-aged adults, 32 compared to NHWs, and these cognitive changes in B/AA adults are associated with smaller changes in CSF tau. 32The ASCEND study included three annual visits (baseline, year

Study design and sample
1, and year 2).Participants underwent LP at baseline and year 2 and a blood draw on all visits.
Medical history and sociodemographic information were collected via a self-report questionnaire.As race is a social construct and collected by self-report, terms and concepts around race and ethnicity are not universal.In the United States some individuals self-identify race as Black, while others identify as African American.To be most inclusive, we are using both terms as they best reflect the racial identities of the study population.

CSF and blood collection and analyses
After an 8 hour overnight fast, participants underwent LP to collect CSF for Aβ and tau, and markers of vascular dysfunction and inflammatory cytokines and chemokines, as previously described. 32Participants also underwent blood draw for analysis of (

Disease-related changes in CSF AD biomarkers over time and according to race
We previously presented cross-sectional baseline measurements of CSF Aβ and tau in these participants. 32Here, we present changes in CSF AD biomarkers over a subsequent 2-year interval from baseline and in relation to race (Table 1); Figure 1 summarizes the major findings.Cut-offs for normal values are based on those established by Gobom et al. 36 as delineated in the Methods section.CSF levels of p-tau increased from baseline to year 2, with moderate effect sizes for the cohort overall (ES = 0.600; P < 0.001), B/AAs (ES = 0.722; P = 0.022), and NHWs (ES = 0.626; P < 0.001).No differences over time for t-tau were found.As previously reported, 32 baseline CSF levels of p-tau (ES = 0.650) and t-tau (ES = 0.853) were lower among B/AA participants than NHW, and these differences remained at year 2 (ES = 0.693 and 0.697, respectively).CSF levels of Aβ 1-40 increased from baseline to year 2 (ES = 0.777; P < 0.001).
As previously reported, 32 baseline CSF levels of Aβ1-40 were lower among B/AA participants than NHW (ES = 0.574) and these differences remained at year 2 (ES = 0.980).No differences by race were found in levels of Aβ1-42.CSF levels of Aβ42/40 were significantly lower at year 2 compared to baseline for the cohort overall (ES = 0.587; P < 0.001), and for B/AAs (ES = 1.184;P = 0.002) and NHWs (ES = 0.722; Participants who were APOE ε4 positive (i.e., having at least one APOE ε4 allele) had lower Aβ1-42 (600 vs. 798 pg/mL; P = 0.001) and Aβ 42/40 (0.068 vs. 0.084; P = 0.0008) compared to those without the ε4 allele.CSF levels of p-tau and t-tau did not differ by APOE ε4 allele when analyzed across the cohort or within groups by race.No associations between AD biomarkers and age or sex were found.
Median p-tau and t-tau levels were within normal range (<50.2 pg/mL and <409 pg/mL, respectively).However, p-tau levels were higher than AD cut-off limits in 12% of participants at baseline and 17% at year 2, while t-tau levels were higher than AD cut-off limits in 26% of participants at baseline and 27% at year 2.

CSF markers of vascular injury differed according to race and were related to markers of neuroinflammation and AD pathology
Mean levels of CSF markers of vascular injury and neuroinflammation at baseline and year 2 are listed in Table 1.There were no differences over time in vascular markers.When stratified according to race, CSF levels of sPDGFRβ were lower among B/AA participants compared to NHW at both baseline (ES = 0.720; P = 0.01) and year 2 (ES = 0.831; P = 0.006), with large effect sizes.In addition, CSF levels of VCAM-1 were lower among B/AA participants compared to NHW at both baseline (ES = 0.734; P = 0.008) and year 2 (ES = 0.970; P = 0.002) with

Circulating levels of vascular markers remain relatively stable
Plasma and serum vascular and inflammatory biomarkers are listed in Table 2.There were no differences in circulating vascular markers from baseline to year 2; however, VCAM-1 was significantly higher at year 1 than baseline (ES = 0.300; P = 0.016).VCAM-1 was significantly lower among B/AAs compared to NHWs at baseline (ES = 0.530) and year 1 (Cohen d = 0.283), with no differences at year 2.

3.5
Circulating levels of markers of neuroinflammation decreased over time IL-10 was significantly lower than baseline at year 1 (ES = 0.309; P = 0.0043) and year 2 (ES = 0.325; P = 0.0058).No significant differences between year 1 and year 2 were found.Among NHWs, IL-10 was significantly lower than baseline at year 1 (ES = 0.315; P = 0.0072) and year 2 (ES = 0.332; P = 0.0135), with no significant differences between year 1 and year 2. No differences over time among B/AAs were found.IL-7 was significantly lower at year 1 than baseline (ES = 0.359; P = 0.0037), and year 2 was significantly lower than year 1 (ES = 0.291; P = 0.0135).No significant differences between baseline and year 2 were found.Among NHWs, IL-7 was significantly lower at year 1 compared to baseline (ES = 0.438; P = 0.029), and year 2 was significantly higher than year 1 (ES = 0.319; P = 0.022), with no differences between baseline and year 2. No differences over time among B/AAs were found.IL-7 was significantly higher among B/AAs compared to NHWs at baseline (ES = 0.541) and year 1 (ES = 0.656), with no differences at year 2.
SAP was significantly lower at year 1 (ES = 0.284; P = 0.0126) and year 2 (ES = 0.233; P = 0.0472) compared to baseline.Among NHWs, SAP was significantly lower than baseline at year 1 (ES = 0.366; P = 0.300), with no significant differences in year 2. No differences between year 1 and year 2 were found.SAP was significantly higher among B/AAs compared to NHWs at all three time points (ES = 0.421, 0.475, and 0.416, respectively).There were no significant differences over time for CRP, a marker of general systemic inflammation.However, CRP was significantly elevated (2-3-fold increase) among B/AA participants compared to NHW at both baseline and year 2, with medium to large effect sizes (ES = 0.757 and 0.442, respectively).IP-10 was collected at year 1 and year 2 only.No differences over time were found among the cohort; however, IP-10 was significantly lower at year 2 compared to year 1 among NHWs (ES = 0.290; P = 0.0272).
No associations between CSF and blood levels of other variables were found.

DISCUSSION
Here we report longitudinal data from a 2-year follow-up period of previously reported baseline results from ASCEND in cognitively unimpaired or mildly impaired B/AA and NHW middle-aged adults with a parental history of AD. 32 Our data demonstrate changes in established AD biomarkers with moderate to large effect sizes, consistent with early disease manifestations.We also highlight significant race-related differences in disease-related and vascular markers within the cohort that require further investigation.
In this cohort of at-risk middle-aged adults, we found changes consistent with increased risk in CSF AD biomarkers between baseline and year 2. While median CSF t-tau remained within normal limits over the 2-year period, there was a significant increase in p-tau from baseline to year 2. CSF Aβ42/40 decreased and Aβ1-40 increased over 2 years.
Although Aβ1-40 is less frequently discussed as a biomarker of AD, several studies have found higher CSF Aβ1-40 in MCI, prodromal AD, and AD; [39][40][41][42] however, other studies found no differences with progression to AD. [43][44][45][46] As this cohort is middle aged with low to moderate vascular risk factors, further work is needed to better understand the contribution of Aβ1-40 to AD risk.
Half of participants at baseline and three fifths of participants at year 2 had below-normal Aβ1-42 levels, indicative of a worsening AD  All bolded values are < 0.05 profile.Lower CSF Aβ1-42 is considered one of the earliest indicators of AD neuropathologic change. 47,48As these findings suggest the higher AD risk inferred by having a first-degree relative with AD may be associated with higher amyloid burden, a more comprehensive understanding of the interplay of family history, genetics, and vascular risk factors among high-risk populations is urgently needed.Ten percent of ASCEND participants met the definition for AD with respect to AT(N) biomarker levels. 2 Together, these data may suggest AD pathological changes in this middle-aged, at-risk cohort, although this cohort is unlikely to have significant evidence of disease.While this cohort has an increased dementia risk due to parental history, the development of dementia is not inevitable.Approximately half of participants in this study were APOE ε4 positive, consistent with other cohorts of adult children of a parent with an AD diagnosis.However, it is important to note we cannot rule out the possibility of parental history of vascular or mixed dementia.We are continuing to follow these individuals over time to better understand the progression of AD biomarkers in high-risk, diverse cohorts.
No changes in CSF vascular markers over 2 years were found.
There were decreases in CSF and plasma markers of inflammation and CSF MMPs from baseline to year 2, which was unexpected considering increasing age and vascular comorbidities are associated with higher vascular dysfunction and inflammation.Changes in these markers over time were not associated with age or comorbidities.Decreases in inflammation could be a product of participation in a research study, as participants had access to education events, which included information on the role of inflammation and risk reduction.
Our baseline findings were the first to report lower levels of t-tau and p-tau based on race in middle-aged adults with an increased risk for AD. 32We now demonstrate tau levels remained lower among B/AA compared to NHW over a 2-year period, consistent with recent analyses of racial differences in AD biomarkers in a population with cognitive impairment. 31,49We also found that CSF markers of vascular dysfunction were lower in B/AA compared to NHW at baseline and year 2.
This supports a recent study in which cardiovascular comorbidities were not associated with CSF tau levels in B/AA participants. 49No association with APOE ε4 was found in this study; this might be explained by the cognitively unimpaired or MCI composition of this cohort. 50 there are differences in AD biomarkers by race, it will be increasingly important to understand appropriate cut-offs for diagnosis and treatments.As potential treatments under development target tau and Aβ, it is important to establish whether certain treatments may be less effective for some communities.Within the research context, there is emerging evidence for differences by race in CSF biomarkers.However, there is insufficient biological data, especially CSF, in racial and ethnic minorities. 51More work is needed globally to better understand these emerging differences along with drivers of AD pathophysiologic processes in diverse populations.
3][54][55][56][57] Higher chronic inflammation among B/AAs has been linked to effects of systemic racism and discrimination, likely due to increased activation of cortisol pathways 56,58,59 and is thought to be a risk factor for cognitive changes and performance on cognitive function assessments in B/AA adults. 53,60However, few studies consider the potential effects of predisposing psychosocial factors, such as stress and discrimination, on the relationship between inflammation and cognitive function.Further work is needed to understand the role of peripheral and central processes in AD risk among diverse populations.
In this cohort of middle-aged adults with a parental history of AD, 36% of participants had one APOE ε4 allele and 11% were homozygous, higher than the general population (9%-23% and 2%-3%, respectively). 61The frequency of APOE ε4 genotype is generally higher among B/AAs (19%) compared to NHWs (14%), 61  increased over time with a decline in MCI compared to those that remained MCI-stable. 63r findings identified mediators of vascular homoeostasis, including MMP-2 and TIMP-1 and TIMP-2 were positively correlated with sPDGFRβ, supporting other work suggesting TNFR dysregulation may be associated with tau pathology and may play a neurodegenerative role in AD pathology. 64,65It is important to note that in a sample size of 81, we are unable to rule out other factors that may mediate relationships between sPDGFRβ and other CSF markers of AD pathology.
Further work in larger studies with diverse populations is needed to better understand the role of sPDGFRβ in early AD pathology.
In conclusion, in a cohort of cognitively unimpaired or mildly impaired middle-aged adults at risk of developing AD, we found evi-

F I G U R E 4
CSF sPDGFRβ was associated with CSF mediators of vascular endothelial injury at year 2. CSF sPDGFRβ was positively associated with CSF TIMP 2 (A), TNFR1 (B), and TNFR2 (C).These measures were not analyzed at baseline.Multiple linear analyses controlling for age, sex, race, and education were used.CSF, cerebrospinal fluid; IL, interleukin; sPDGFRβ, soluble platelet-derived growth factor beta; TIMP, tissue inhibitor of metalloproteinase; TNFR, tumor necrosis factor receptor; VCAM-1, vascular cell adhesion molecule-1.
although there were no differences in this study.CSF Aβ42 and Aβ42/40 were significantly lower in persons with at least one APOE ε4 allele, which may relate to additional AD risk in persons with a parent with AD.While APOE ε4 genotype confers an increased genetic risk factor, not all persons who are ε4 positive develop AD, even with a family history.Considering other known risk factors, many of which are vascular, other possible outcomes include the development of vascular dementia or a mixed phenotype.We also report associations between CSF markers of vascular injury and neuroinflammation in relation to markers of established disease pathology, indicative of early disease changes.CSF sPDGFRβ was positively associated with markers and mediators of vascular injury and brain cytokines and correlated with disease pathology.Interestingly, CSF sPDGFRβ, ICAM-1, and ACE-1 were lower in B/AA compared to NHW for reasons that are not yet clear.These findings are surprising considering the higher proportion of B/AAs with a history of hypertension compared to NHWs in this cohort.These data suggest vascular injury and neuroinflammation are related to early changes in AD pathology in a cognitively unimpaired or mildly impaired middle-aged at-risk cohort.Markers of vascular injury, particularly sPDGFRβ, were strongly associated with levels of p-tau, t-tau, and Aβ1-40, but surprisingly not Aβ1-42.However, persons with low Aβ1-42 had a larger increase in sPDGFRβ compared to individuals with normal Aβ1-42, supporting a previous study showing CSF sPDGFRβ level correlated with t-tau and p-tau in AD. 25 These data suggest pericyte dysfunction occurs early in the disease process, but further studies are required to confirm whether CSF sPDGFRβ changes are related to, or independent from, changes in CSF tau and Aβ.Our data indicate sPDGFRβ is associated with markers of vascular endothelial activation, mediators of endothelial dysfunction, and IL-9.These findings support previous observations linking vascular dysfunction and central nervous system inflammation in the early presymptomatic phases of dementia.In a study of persons with MCI and AD, raised CSF levels of vascular markers and cytokines were associated with t-tau and p-tau, cortical thinning, and poorer Mini-Mental State Examination scores; the associations were stronger in Aβ-positive individuals. 11Pericyte loss and BBB breakdown have been associated with elevated levels of brain cytokines in early stages of AD, 62 and serum markers of vascular injury and inflammation dence of changes in established CSF biomarkers at 2-year follow-up consistent with early changes in the development of AD (CSF ttau/p-tau and Aβ40).CSF markers of vascular dysfunction, including pericyte damage and endothelial injury, were associated with changes in MMP-2 and TIMPs, and cytokines.Elevated sPDGFRβ was related to t-tau, p-tau, and changes in Aβ40 but not Aβ42.Interstitial fluid tau, sPDGFRβ, ACE-1, and ICAM-1 were lower in B/AA compared to NHW, which is worth noting with respect to what are perceived normative ranges for patients of different racial backgrounds.These data indicate that cerebral vascular dysregulation may be a very early event that occurs in the development of AD, detectable in a mid-life cognitively unimpaired or mildly impaired cohort, and that race impacts these relationships, warranting further research.
33seline results were reported previously.32Briefly,weenrolled82 middle-aged (≥45 years) adults with a biological parent with either autopsy-confirmed or probable AD as defined by National Institute of Neurological and Communicative Disorders and Stroke and Alzheimer's Disease and Related Disorders Association criteria33and verified using the validated Dementia Questionnaire 34 and med- The Association Between Cardiovascular Risk and Preclinical Alzheimer's Disease Pathology (ASCEND) Study was a 2-year observational study of cognitively unimpaired or mildly impaired, middle-aged B/AA and NHW adults at risk for AD due to parental history.used to assess cognitive impairment.MoCA scores >26 were considered normal, and scores between 18 and 25 were categorized as MCI.
TA B L E 1 CSF biomarkers by race over 2 years.
Plasma/serum biomarkers by race over 2 years.
TA B L E 2 Note: Results are reported as mean ± standard deviation or median (95% CI).Two-sample t-tests for normally distributed data and paired Wilcoxon signed rank test for non-normally distributed data were used to test differences by race at each time point.Least squares mean analyses were used to test differences over time, differences by race at each time point.Abbreviations: ACE, angiotensin-converting enzyme; CI, confidence interval; CRP, C-reactive protein; ICAM, intercellular adhesion molecule; IFN, interferon; IL, interleukin; IP, interferon gamma-induced protein; MCP, monocyte chemoattractant protein; MDC, macrophage-derived chemokine; MMP, matrix metalloproteinase; SAP, serum amyloid protein; TGF, tumor growth factor; TNF, tumor necrosis factor; TNFR, tumor necrosis factor receptor; VCAM, vascular cell adhesion molecule.*P-values represent differences by race at each time point.